Content uploaded by Daniel A. Smith
Author content
All content in this area was uploaded by Daniel A. Smith on Mar 21, 2014
Content may be subject to copyright.
78
State and Local Government Review
Vol. 38, No. 2 (2006): 78–91
D
banning same-
sex
marriage swing the 2004 gen-
eral election to George W. Bush?
In 2004, activists and state legislators placed
anti-gay marriage questions on the general
election ballots of 11 states. All of the ballot
measures passed easily, receiving on average
roughly 70 percent support.
1
Pundits argued
that the marriage measures on the November
ballot would be a major motivating factor in
the election and would help ensure Bush’s
reelection. The measures, so the logic went,
would receive broad support from social con-
servatives who would be mobilized to go to
the polls primed to vote for Bush, who was
rmly aligned with the issue.
Merging county-level religious, socioeco-
nomic, and political data with the 2004 elec-
tion results, this article examines the electoral
effects of same-sex marriage ballot measures
in two key neighboring presidential battle-
ground states, Michigan and Ohio. In both
states, the anti-gay marriage measures passed
easily, and overall statewide turnout increased
from the 2000 election. However, Bush pre-
vailed only in Ohio. To test the impact of the
anti-gay marriage measures in both states, the
analysis proceeds in three stages, examining
rst the county-level voting patterns on the
gay marriage bans, then patterns of county-
level turnout in 2004, and nally the county-
level support for President Bush in 2004.
The Mobilizing and
Priming Effects of Ballot Measures
Following the election, numerous opinion-
makers claimed that the anti-gay marriage
measures—especially the initiative on the
ballot in the swing state of Ohio—were
the key to Bush’s win (Davies 2004). Tony
Perkins, president of the Family Research
Council, asserted after the election that gay
marriage was “the hood ornament on the
family values wagon that carried the presi-
dent to a second term” (Cooperman and
Edsall 2004, A-1). The director of the con-
servative Culture and Family Institute think
tank, Robert Knight, claimed the anti-gay
marriage measures “galvanized millions of
Christians to turnout and vote, and George
Bush and the GOP got the lion’s share of
that vote” (Lochhead 2004, A-1). For his
part, Karl Rove, President Bush’s principal
advisor, told reporters at the Republican Na-
tional Convention in August 2004 that “to
the degree it energizes people who might
otherwise not vote,” a ballot measure ban-
ning gay marriage “tends to help us” (quoted
in Korte 2004, A-1).
Same-Sex Marriage Ballot Measures
and the 2004 Presidential Election
Daniel A. Smith, Matthew DeSantis, and Jason Kassel
79
Vol. 38, No. 2, 2006
Same-Sex Marriage Ballot Measures
After the election, reporters for many na-
tional and local papers, including the vener-
able New York Times, reported matter-of-factly
that Issue 1 “helped turn out thousands of
conservative voters on Election Day” and
that support for the measure was “widely
viewed as having been crucial to President
Bush’s narrow victory in that swing state”
(Dao 2004, A-28). Opining in his postelec-
tion USA Today syndicated column that “gay
marriage was probably the issue that spelled
the difference in hard-fought Ohio” and that
Bush “directly bene ted from the outpour-
ing of social conservatives” who turned out
in support of Issue 1, reporter Walter Shapiro
(2004, A-9) concurred with an Ohio Demo-
cratic Party strategist who claimed that “if
Issue 1 had not been on the ballot, John Kerry
would have won Ohio.” A postelection analy-
sis of the county-by-county presidential vote
conducted by the Wall Street Journal found
that “Bush won thanks to a pitch on mor-
als that went beyond evangelicals to Roman
Catholics” and “a strong effort to turn out
rural voters” (Cummings 2004, A-4). Bush did
particularly well, the paper reported, in small
and conservative counties, including 12 of 14
with the most heavy job losses since 2000. A
story in USA Today reinforced this conven-
tional wisdom that smaller, more conservative
exurban, or “micropolitan,” counties in Ohio
were central to Bush’s victory, with the Re-
publicans winning 27 of 29 such counties with
“rural sensibilities” (El Nasser 2004, A-1).
In contrast to the conventional wisdom of
the media following the election, recent schol-
arship on the electoral effects of anti-gay mar-
riage amendments is decidedly mixed. Relying
on aggregate national turnout data and state-
wide presidential votes, Abramowitz (2004)
and Burden (2004) found that the 11 states
with gay marriage bans in November did not
have higher voter turnout or support for Bush
than states without the bans (after control-
ling for a state’s rate of turnout in 2000) if a
state had a 2004 U.S. Senate race or if it was
a “swing” state in 2004. By extension, they
contend that same-sex marriage bans did not
contribute to the reelection of Bush. Prelimi-
nary analyses using county-level data are more
mixed. A few indicate that support for the gay
marriage bans did not lead to higher turnout
or a positive effect of support for Bush in Ohio
(Jackman 2004; Freedman 2004), but others
suggest that the measures may have in uenced
Bush’s victory across the 11 states with the
gay marriage bans, though they did not boost
turnout (Campbell and Monson 2005). Stud-
ies using individual-level data national surveys
are also mixed. Some conclude that the issue
of “moral values” (including support for ban-
ning gay marriage) was not central to Bush’s
reelection (Hillygus and Shields 2005; Lewis
2005). Others, though, nd that the issue of
gay marriage was more important to voters
living in states where the measures were on
the ballot and, more signi cantly, may have
helped to prime voters’ support for Bush
(Donovan, Tolbert, and Smith 2005).
Theoretically, there are good reasons why
ballot measures—particularly ones so polar-
izing as those banning same-sex marriage—
might be expected to mobilize certain citizens
to go to the polls and prime them to vote for
candidates. First, there is solid scholarly evi-
dence supporting the assumptions about the
potential electoral effects of ballot measures
(Smith and Tolbert 2004; Nicholson 2005).
Research has shown that ballot measures
increase voter turnout, particularly among
partisans (Smith 2001; Tolbert, Grummel,
and Smith 2001; Tolbert and Smith 2005).
Moreover, voters tend to draw on elite and
partisan cues to inform their decisions on how
to cast their ballots (Lupia 1994; Bowler and
Donovan 1998; Lewkowicz 2006), and the
ensuing votes tend to be split along partisan
lines (Branton 2003; Smith and Tolbert 2001).
Recent scholarly evidence has shown that
ballot measures may have priming effects on
candidate races (Nicholson 2005; Donovan,
Tolbert, and Smith 2005).
The Political Context
In the buildup to the 2004 election, there were
many reasons to expect that gay marriage
80
Smith, DeSantis, and Kassel
State and Local Government Review
amendments might create a perfect storm in
the reelection of the president. In late Febru-
ary 2004, the president expressed his support
for a U.S. constitutional amendment banning
gay marriage. Citing the efforts of activist
judges in Massachusetts and local of cials
in San Francisco to rede ne marriage, Bush
claimed that the 1996 Defense of Marriage
Act (DOMA), signed into law by President
Bill Clinton, could be undermined by the
courts and did not protect marriage in the
states. Nearly 40 states already had their own
versions of DOMA on the books, including
both Michigan and Ohio (Williams 2004).
Over the summer, though, the U.S. Senate
rejected a federal marriage constitutional
amendment, which many media outlets re-
ported as “a big election-year defeat” for the
White House (Cooperman and Edsall 2004,
A-1). Meanwhile, in roughly a dozen states
already having statutory DOMAs, including
Michigan and Ohio, there were efforts under
way in early 2004 to place amendments on
statewide ballots that would constitution-
ally de ne marriage as between a man and a
woman (O’Connell 2006).
By April, proponents in both Ohio and
Michigan announced they would begin cir-
culating petitions to place amendments with
traditional de nitions of marriage on the
November ballots. In each state, the gay
and lesbian community announced a coun-
termobilization strategy and initiated court
challenges to both the validity of the peti-
tions’ signatures and the amendments’ con-
stitutional language. State courts, though,
rejected these challenges, and the measures
were subsequently placed on the ballot for
citizen approval.
In early August, the Ohio Campaign to
Protect Marriage (OCPM) submitted 575,000
signatures to the Ohio Secretary of State, eas-
ily surpassing the 316,888 valid signatures
needed to place a constitutional amendment
on the ballot, and the Ohio Ballot Board
unanimously approved the measure’s title and
two-sentence ballot summary. Issue 1 would
amend Section 11 of Article XV of the state
constitution, stipulating that “only a union
between one man and one woman may be a
marriage valid in or recognized by this state
and its political subdivisions” and that the
“state and its political subdivision shall not
create or recognize a legal status for relation-
ships of unmarried individuals that intends to
approximate the design, qualities, signi cance
or effect of marriage.” On September 29, the
Ohio Secretary of State certi ed the initiative
for placement on the ballot.
In Michigan, the State Court of Appeals
announced in September that the more than
317,000 signatures gathered by Citizens for
the Protection of Marriage and submitted in
July for veri cation were valid, which secured
the measure’s spot on the general election bal-
lot. The court ruling was necessary after the
Michigan Board of State Canvassers was dead-
locked in its decision to place the issue before
the voters. The court ruled that the proposed
constitutional amendment’s stipulation that
the state would only recognize a union be-
tween a man and woman as a marriage or
“similar union for any purpose” was neither
incomplete nor misleading. As in Ohio, critics
of the measure (dubbed Proposal 2) had ar-
gued that the ballot language was overly broad
and could force universities, governments, and
other public agencies to stop giving domestic
partner health bene ts to same-sex couples
and heterosexual couples who were not mar-
ried. The appellate court ruled that the mem-
bers of the Board of State Canvassers were
wrong to consider the substantive merits of
the proposal and that its legality could not be
challenged unless it became law.
Research Design
Two separate aggregate-level datasets for
the 88 counties in Ohio and 83 counties in
Michigan are used to probe where the ballot
measures banning same-sex marriage gained
electoral support and to test whether they
helped increase levels of turnout and aid in
the reelection of Bush across the counties
in each state. Although aggregate-level data
81
Vol. 38, No. 2, 2006
Same-Sex Marriage Ballot Measures
cannot be used to infer individual patterns of
political behavior without risking an ecologi-
cal fallacy (King 1997), county-level data are
useful when exploring the political, economic,
and social environment in which voters make
their electoral decisions (Key 1949; Giles
1977; Oliver and Mendelberg 2000; Hero
1998; Tolbert and Hero 2001). Aggregate-
level data are especially useful when exploring
turnout and voting patterns, as survey data
may contain nonresponse and other sampling
errors or may accentuate the social desirability
of participation and vote choice (Belli et al.
1999). In both Michigan and Ohio, counties
are particularly useful units of analysis because
they are the administrative arms of the state
and are key electoral arenas in which candi-
dates and parties wage their campaigns.
Previous research employing aggregate-
level analyses of political behavior demon-
strates that partisan, religious, and socio-
economic compositions often in uence the
overall political behavior of communities or
counties (Putnam 1966; Foladare 1968; Huck-
feldt, Plutzer, and Sprague 1993). For example,
Foladare (1968) demonstrated that Protestants
who lived in predominantly Catholic commu-
nities had signi cantly different partisan iden-
ti cation than those Protestants who lived in
communities that had a large Protestant pop-
ulation, and vice-versa. Paralleling religious
identi cation, scholars also have demonstrated
that the racial composition of a state in uences
the type of racial policies it enacts (Giles and
Evans 1986; Hero and Tolbert 1996; Hero
1998; Branton 2004). The research design
combines elements from these perspectives
in assessing the in uence of various communi-
ties on the voting patterns on the gay marriage
issue and changes in the level of turnout and
the president’s vote share in each state.
Expectations and Speci cations
of Models
This investigation into the electoral and turn-
out effects of the anti-gay ballot measures
lends itself to three testable hypotheses.
2
H1: Counties with greater densities
of Republicans, rural and exur-
ban populations, and evangeli-
cal Protestants and Catholics will
have higher levels of support for
the ballot measures banning same-
sex marriage compared with other
counties.
H2: Counties with greater support for
the measures banning same-sex
marriage will have higher levels
of turnout in 2004 compared with
other counties.
H3: Counties with greater support for
the measures banning same-sex
marriage will have higher levels of
support for Bush in 2004 compared
with other counties.
The rst hypothesis provides insight into the
forces at play in the passage of the anti-gay
measures, while the second and third hypoth-
eses capture the impact the gay marriage bans
may have had on turnout and vote for Bush
in 2004. A series of weighted least-squares
multivariate regression models estimate the
aggregate-level patterns of voting on the bal-
lot issues, turnout, and support for Bush in
Michigan and Ohio in 2004, controlling for
a host of other factors.
3
Across Michigan’s 83
counties, turnout in 2004 was on average 4.9
percentage points greater than in 2000, and
the vote for Bush increased 1.8 percentage
points from 2000 to 2004. In Ohio, turnout
across the state’s 88 counties was 8.2 per-
centage points higher in 2004 than in 2000,
and Bush’s vote share increased 2 percentage
points after his four years in of ce. Table 1
details the statewide averages and the county
averages of the vote for same-sex marriage
bans, the percentage point change in voter
turnout, and the percentage point change in
vote for Bush.
Expected Support for the
Gay Marriage Bans
The rst model tests whether counties in
Ohio and Michigan with greater densities of
82
Smith, DeSantis, and Kassel
State and Local Government Review
religious conservatives, Republicans, and rural
and exurban populations exhibited increased
levels of support for the initiatives banning
gay marriage, holding other factors constant.
The dependent variable is the percent vote in
favor of Issue 1 in each of Ohio’s 88 counties
and the percent vote in favor of Proposal 2
in Michigan’s 83 counties.
Counties with dense populations of evan-
gelical Protestants are expected to have
stronger support for the measures in the
two states. Evangelicals tend to view same-
sex ballot measures, which restrict the rights
of gays, as a tactic to battle against a way of
life they perceive as threatening to their own
social and cultural values (Button, Rienzo,
and Wald 1997; Wald, Owen, and Hill 1990).
There is also evidence at the aggregate level
that communities with higher percentages
of evangelicals are likely to adopt ordinances
limiting the rights of gays and lesbians (But-
ton, Rienzo, and Wald 1997, 80–82). There
is therefore good reason to expect that the
ballot measures banning gay marriage will
have higher levels of support in counties with
large percentages of evangelicals.
The model includes the county popula-
tion in each state belonging to the Catholic
Church. While not so evident in Ohio, where
the backing for Issue 1 came principally from
the Ohio Campaign to Protect Marriage, in
Michigan the Catholic Church actively cam-
paigned and nancially supported Citizens
for the Protection of Marriage, the primary
backers of Proposal 2. In mid-October, for
instance, the Michigan Catholic Confer-
ence mailed a letter and a brochure touting
marriage between a man and a woman to all
596,000 households in the state with a regis-
tered Catholic. Seven Catholic dioceses (led
by the Archdiocese of Detroit) contributed
$1 million to Citizens for the Protection
of Marriage, accounting for 61 percent of
the war chest accumulated by the primary
umbrella group advancing the amendment
(O’Connell 2006; Cooperman and Broder
2004). In Michigan (but not Ohio), due to
the Catholic Church’s heavy involvement in
the campaign, the percentage of Catholics
in a county is expected to have a positive ef-
fect on county-level support for the initiative.
The model also controls for the percentage
of a county’s population belonging to other
denominations that did not of cially oppose
the gay marriage bans. As with Catholics and
evangelicals, the expectation is that as the
density of adherents belonging to conserva-
tive denominations increases in a county, so
too will the support for the measures ban-
ning gay marriage. By default, the excluded
(reference) category is the percentage of a
county’s population belonging to liberal de-
nominations, combined with the percentage
of a county’s population with no of cial reli-
gious adherence.
4
Backers of the measures also hoped that
the initiatives would mobilize socially con-
servative Republicans, priming their support
for Bush. Partisanship has been shown to be
the most predicable indicator of support or
opposition for ballot measures (Smith and
Tolbert 2001; Branton 2003). In 2004, the
issue of gay marriage was de ned along par-
tisan lines, with the Republican Party gener-
ally supporting (if sometimes tacitly) the ban
and Democrats generally opposing it. It is
expected that there will be greater support for
the ballot initiatives in counties with larger
Table 1. Statewide and County Averages
Michigan Ohio
Variables Statewide County Statewide County
Vote in favor of initiative banning gay marriage (percent) 58.6 63.8 61.7 68.5
Percentage point change in turnout, 2000–2004 5.7 4.9 8.2 8.2
Percentage point change in vote for Bush, 2000–2004 1.7 1.8 0.8 2.0
83
Vol. 38, No. 2, 2006
Same-Sex Marriage Ballot Measures
percentages of Republicans; partisanship is
therefore controlled with a measure of the
county percent vote for Bush in 2000 in both
states, as neither state of cially records the
party af liation of registered voters.
5
Propo-
nents of the measures also targeted support
from voters living in rural and exurban coun-
ties. Using a 9-point scale created by the U.S.
Department of Agriculture (ranging from
1 = metropolitan area of 1 million population,
to 9 = completely rural population in juris-
dictions with fewer than 2,500 residents) to
control for the rural/urban nature of a county,
the expectation is that more rural counties
would exhibit greater levels of support for
same-sex marriage bans.
Also included in the model are traditional
socioeconomic control variables that research-
ers have found to be signi cant predictors of
support or opposition to gay rights measures
(Button, Rienzo, and Wald 1997; Donovan
and Bowler 1998). Advocates of the measures
also hoped to garner the support of socially
conservative African Americans, envisioning
that the wedge issue would pry them away
from the Democratic Party’s candidate. While
many African Americans self-identify as evan-
gelicals and have a theology that often re ects
that of the Christian Right, their voting habits
tend to be de ned by liberal economic issues
(Wilcox 2000). As such, counties with dense
African American populations, measured by
the percentage of African Americans residing
in each county in 2000, are likely to be cued
by the Democratic Party to vote against the
anti-gay marriage initiatives.
Educational attainment and age have been
found to be major cleavages in the larger gay
rights debate; individuals with lower levels of
education and older individuals generally op-
pose extending the rights of gays and lesbians
(Button, Rienzo, and Wald 1997). Counties
with a higher median age and lower level of
education (as measured by the percentage of
the county’s population over 25 years old with
at least a high school education) are expected
to have greater support for the bans on same-
sex marriage.
Finally, the economic well-being of a
county is controlled with a measure of the
percent change in a county’s rate of unem-
ployment between 2000 and 2003.
6
Many
counties in both Ohio and Michigan were
hit hard by rising unemployment levels dur-
ing President Bush’s rst term. Organized
labor in these two heavily unionized states
decided to stay on the sidelines in the pitched
battle over same-sex marriage, focusing at-
tention instead on economic issues. Citizens
in counties with increasing rates of employ-
ment may have grown frustrated with the
Bush campaign’s symbolic crusade on moral
issues. Motivated instead by job growth and
more immediate material relief, counties ex-
periencing increasing levels of unemployment
between 2000 and 2003, controlling for other
factors, are expected to have less support for
the anti-gay marriage initiatives relative to
other counties.
Expected Voter Turnout in 2004
The second model estimates the possibility of
a mobilization effect of the two ballot mea-
sures. The dependent variable is the percent
turnout in each of Michigan’s 83 and Ohio’s
88 counties in 2004. As mentioned previously,
supporters of same-sex marriage bans claimed
that the measures on the ballot would mo-
bilize conservative evangelicals, Catholics,
and rural voters. Most notably, evangelicals
were considered to be a bloc of voters who
had allegedly stayed home four years earlier
(see Teixeira 2004). Previous studies nd that
the mobilization of evangelical Protestants
is maximized when issues such as abortion,
the rights of gays and lesbians, and school
prayer are either on the ballot or are major
components of a political campaign (Wilcox
1989; Jelen 1991; Rozell and Wilcox 1996;
Carmines and Layman 1997; Wilcox 2000),
and gay rights ballot measures are often seen
as polarizing issues, with clear partisan cleav-
ages (Donovan, Wenzel, and Bowler 2000;
Donovan and Bowler 1998; Witt and Mc-
Corkle 1997). Of course, the strategy to mo-
bilize evangelicals in support of the anti-gay
84
Smith, DeSantis, and Kassel
State and Local Government Review
marriage measures was vulnerable to counter-
mobilization strategies (Reichley 1986; Green
1994; Rozell and Wilcox 1996; Haider-Markel
and Meier 1996; Wilcox 2000). In estimating
the 2004 county-level turnout in each state,
the model uses the same control variables as in
the previous model but adds percent turnout
in 2000 and the percent vote in favor of the
anti-gay measures in 2004. Both variables are
expected to have positive effects on county-
level turnout in 2004, all else held equal.
Expected Vote Share for Bush in 2004
The nal model, also essentially a panel model,
estimates the effect same-sex ballot measures
had on presidential vote choice in 2004 in
each state. The dependent variable is Bush’s
share of the county-level presidential vote in
2004. The main control variable estimating
the vote for Bush in 2004 is the level of sup-
port Bush received in 2000 and is expected to
be strongly positive. Even though evangeli-
cal Protestants are not always a cohesive vot-
ing bloc and are sometimes apolitical (Smidt
1988; Wilcox 1989; Jelen and Wilcox 1992;
Green 1994; Rozell and Wilcox 1996; Shibley
1998; Wilcox 2000), the density of evangeli-
cals in a county is expected in Ohio to help
explain support for Bush, as Issue 1 was vigor-
ously pushed by evangelical leaders. Similarly,
in Michigan, the county density of Catholics
is expected to be positively related to the vote
for Bush. The other independent variables are
identical to those in the previous model, with
the same expected directions.
Findings
The results of the weighted least-squares
regressions for both states are presented in
Tables 2–4. What is most striking about the
results is that support for anti-gay marriage
measures in both states affected the vote for
Bush in 2004, even when controlling for his
level of support in 2000. A county’s percentage
of evangelicals was also positively signi cant
in Ohio (but not in Michigan) when estimat-
ing support for Bush in 2004. County-level
turnout in 2004, however, was not affected
by the level of support for same-sex marriage
measures in Ohio and was actually negatively
related in Michigan. In sum, counties with
higher levels of support for the measures
banning same-sex marriage appear to have
had greater support for Bush in 2004, though
not higher turnout, compared with the 2000
election.
Support for the Gay Marriage Bans
Consistent with scholarly research show-
ing partisanship at both the individual and
aggregate levels to be a strong predictor of
vote choice on ballot measures (Branton 2003;
Smith and Tolbert 2001), anti-gay marriage
measures in both states received greater sup-
port (as measured by the percent vote for Bush
in 2000) in counties with greater percentages
of Republicans (see Table 2). In Michigan, for
every percentage point increase in a county’s
support for Bush in 2000, support for Pro-
posal 2 increased by nearly half a percentage
point (.48). Similarly, in Ohio, support for
Issue 1 increased by a third of a percentage
point (.31) for every percentage point increase
in a county’s support for Bush in 2000.
The results also indicate that in both Mich-
igan and Ohio, the density of evangelicals was
positively related to county vote in support of
gay marriage bans. Stated differently, support
for the two ballot measures in counties with
denser populations of evangelical Protestants
was more than in counties with less dense
populations, ceteris paribus. In both Michigan
and Ohio, support for the anti-gay marriage
measures rose by almost a quarter of a per-
centage point for every one point increase in
a county’s percentage of evangelicals. Fur-
thermore, consistent with research on gay
rights ballot issues showing that levels of tol-
erance are commensurate with educational
attainment (Button, Rienzo, and Wald 1997;
Donovan and Bowler 1998), as educational
attainment increased, the vote for the gay
marriage ban decreased in both states (more
so in Michigan than Ohio). Finally, control-
ling for other factors, the Ohio model shows
85
Vol. 38, No. 2, 2006
Same-Sex Marriage Ballot Measures
Table 2. Percent Vote for Initiatives Banning Same-Sex Marriage, 2004
Michigan Ohio
County-level Variables (percent vote for Proposal 2) (percent vote for Issue 1)
Percent evangelical Protestant, 2000 .231 (.098) .245 (.110)
Percent Catholic, 2000 .084 (.041) .000 (.036)
Percent other conservative denominations, 2000 .137 (.112) .072 (.082)
Percent African American, 2000 −.031 (.047) −.257 (.049)
Urban/rural continuum, 2003 (1 = urban; 9 = rural) −.002 (.002) .006 (.003)
Median age, 2000 .001 (.001) .012 (.002)
Change in percent unemployment, 2000–2003 −.075 (.369) −.778 (.401)
Percent population with high school degree, 2000 −.864 (.092) −.511 (.101)
Percent vote for Bush, 2000 .482 (.057) .312 (.041)
Constant .969 (.102) .468 (.114)
Adjusted R
2
.868 .678
F 61.170 21.360
N 83 88
Table 3. Percent Turnout, 2004
County-level Variables Michigan Ohio
Percent vote for same-sex marriage bans −.092 (.048) −.220 (.156)
Percent evangelical Protestant, 2000 .056 (.041) .087 (.156)
Percent Catholic, 2000 .006 (.018) .008 (.050)
Percent other conservative denominations, 2000 .013 (.046) −.029 (.112)
Percent vote for Bush, 2000 .084 (.036) .089 (.083)
Percent voter turnout, 2000 .861 (.052) .594 (.153)
Percent African American, 2000 −.029 (.019) −.079 (.079)
Median age, 2000 .000 (.000) .010 (.003)
Urban/rural continuum, 2003 (1 = urban; 9 = rural) −.003 (.001) .000 (.004)
Percent population with high school degree, 2000 .067 (.057) −.108 (.164)
Change in percent unemployment, 2000–2003 .253 (.149) −.835 (.561)
Constant .089 (.062) .199 (.173)
Adjusted R
2
.959 .402
F 176.748 6.323
N 83 88
*p < .10.
Notes: Least-squares regression is weighted by total population of county. Each model reports unstandardized regression
coeffi cients; standard errors are in parentheses. Bold coeffi cients indicate observed statistical reliability at 90 percent con-
fi dence intervals.
Sources: U.S. Census Bureau, 2000; Ohio Secretary of State, 2004; Michigan Secretary of State, 2004; U.S. Department
of Agriculture, 2004; Glenmary Research Center (Jones et al. 2002); Ohio Department of Job and Family Services, Bureau
of Labor Market Information, 2004; Michigan Offi ce of Labor Market Information, Bureau of Labor and Economic Growth
& Strategic Initiatives, 2004.
signi cant variables, but none were found in
either state. The Ohio model indicates that
the density of African Americans in a county
was signi cant; controlling for other factors:
as the percentage of African Americans re-
siding in a county increased, support for Is-
sue 1 decreased. This nding appears to run
that counties experiencing poor economic
conditions exhibited less support for Issue 1
relative to other counties.
In addition to the independent variables
that help to explain a county’s level of sup-
port for the gay marriage bans in both Michi-
gan and Ohio, there were a handful of other
86
Smith, DeSantis, and Kassel
State and Local Government Review
counter to individual-level survey data, which
show that nonwhites in Ohio (but not in
Michigan) were more likely to support Issue
1 (Donovan, Tolbert, and Smith 2005). The
Ohio model shows that, as expected, more
rural counties and those with older popula-
tions were signi cantly more likely to support
the ban and those experiencing higher levels
of unemployment were less likely to support
the ban . In Michigan, the model shows that
the density of Catholics in a county was a
signi cant predictor of support for Proposal
2, as anticipated. For each percentage point
increase in a county’s density of Catholics,
support for Proposal 2 increased by a little
less than a tenth of a percentage point, hold-
ing constant other factors. In both states, the
percentage of other conservative denomina-
tions in a county had no signi cant impact on
level of support for the measures.
Voter Turnout in 2004
The outcome variable in Table 3 is percent
turnout in a county in 2004. Controlling for
several factors (most notably, county-level
turnout in 2000), the Ohio model shows that
support for the gay marriage ban had no sta-
tistically signi cant impact on county turnout
in 2004. In Michigan, the anti-gay marriage
measure had a slight negative effect: for every
percentage point increase in county support
for Proposal 2, turnout was less by one-tenth
of a percent. Furthermore, in neither state
was the percentage of evangelicals related to
increased levels of turnout in 2004. These
ndings indicate that the higher county-
level turnout in both states in 2004 was not
due to the support for the anti-gay measures
or to those counties with dense evangelical
populations having heavier turnout. Rather,
in both states, turnout in 2004 was up across
all counties, irrespective of the aggregate
religious persuasion of a county or whether
the median voter in a county supported the
anti-gay marriage amendment. Perhaps sur-
prisingly, there is no single common explana-
tory factor in the two states for the increase
in turnout. In Michigan, the model shows
Table 4. Percent Vote for Bush, 2004
County-level Variables Michigan Ohio
ß (se) ß (se)
Percent vote for same-sex marriage bans .269 (.047) .212 (.053)
Percent evangelical Protestant, 2000 −.045 (.041) .105 (.053)
Percent Catholic, 2000 −.057 (.017) .015 (.017)
Percent other conservative denominations, 2000 −.172 (.046) −.019 (.038)
Percent vote for Bush, 2000 .775 (.032) .957 (.025)
Percent African American, 2000 −.102 (.019) −.038 (.027)
Median age, 2000 .001 (.000) .001 (.001)
Urban/rural continuum, 2003 (1 = urban; 9 = rural) −.001 (.001) .000 (.001)
Percent population with high school degree, 2000 .169 (.055) .054 (.054)
Change in percent unemployment, 2000–2003 .175 (.149) −.068 (.191)
Constant −.158 (.062) −.180 (.059)
Adjusted R
2
.992 .989
F 1073.604 797.993
N 83 88
*p < .10.
Notes: Least-squares regression is weighted by total population of county. Each model reports unstandardized regression
coeffi cients; standard errors are in parentheses. Bold coeffi cients indicate observed statistical reliability at 90 percent con-
fi dence intervals.
Sources: U.S. Census Bureau, 2000; Ohio Secretary of State, 2004; Michigan Secretary of State, 2004; U.S. Department
of Agriculture, 2004; Glenmary Research Center (Jones et al. 2002); Ohio Department of Job and Family Services, Bureau
of Labor Market Information, 2004; Michigan Offi ce of Labor Market Information, Bureau of Labor and Economic Growth
& Strategic Initiatives, 2004.
87
Vol. 38, No. 2, 2006
Same-Sex Marriage Ballot Measures
that more rural counties had lower levels of
turnout, whereas counties with higher levels
of unemployment had higher turnout com-
pared with other counties. In Ohio, counties
with older populations had higher turnout in
2004, controlling for other factors.
Vote Share for Bush in 2004
The results shown in Table 4 provide con-
siderable evidence that support in both states
for same-sex marriage ballot measures led
to increased levels of support for Bush and
that, in Ohio, counties with dense evangelical
populations contributed to Bush’s reelection.
Consistent with survey research showing that
the gay marriage bans had a priming effect for
Bush in 2004 (Donovan, Tolbert, and Smith
2005), county-level support for Ohio’s Issue
1 and Michigan’s Proposal 2 appears to have
given the president a boost. In Ohio, con-
trolling for the level of support the president
received in 2000 and other factors, the vote
for Bush in 2004 was positively related to
support for the ban on gay marriage, as was
the density of a county’s evangelical popula-
tion. In Ohio—a state where Bush ended up
winning by only 118,000 votes in 2004—the
model reveals that for every percentage point
increase in the vote for Issue 1, support for
Bush increased by two-tenths of a percentage
point; support for Bush increased an addi-
tional tenth of a percentage point for every
percentage point increase in density of evan-
gelicals in a county. Thus, while there is no
evidence that counties in Ohio with denser
evangelical populations and support for Issue
1 had higher turnout (Table 3), they did pro-
vide more support for Bush in 2004 relative
to other counties.
7
As in Ohio, county-level support in Michi-
gan for the proposal to ban same-sex marriage
positively affected the vote for Bush in 2004
(see Table 4). For every percentage point in-
crease in the vote for the anti-gay marriage
measure, county-level support for Bush went
up by a quarter of a percentage point. Bush
also received stronger support in counties
with more educated populations, controlling
for other variables. However, Bush’s support
in the state was offset by several other factors.
Unlike in Ohio, the county-level density of
evangelicals in Michigan was not signi cant
and does not help to explain county vote share
for Bush in 2004. Additionally, the coef cients
for the density of Catholics and the percent-
ages of African Americans and adherents of
other conservative denominations in a county
all were signi cant but in a negative direction,
indicating that these counties had lower levels
of support for Bush in 2004 relative to other
counties.
Conclusion
The county-level ndings from the 2004
battleground states of Michigan and Ohio
support but also raise questions about the
electoral impact of same-sex marriage bans.
There is evidence that counties with dense
levels of evangelical Protestants in both
states and counties with sizeable numbers
of Catholics in Michigan voted strongly in
favor of the anti-gay ballot measures. Even
more impressive was the positive correla-
tion in both states between support for the
anti-gay marriage measures and the vote for
Bush in 2004, even after taking into consid-
eration the strong predictive value of the vote
for Bush in 2000. Additionally, in Ohio the
percentage of evangelicals in a county was a
positive predictor of the vote for Bush. Any
gains for Bush from the same-sex marriage
measure in Michigan, however, were seem-
ingly countered by signi cantly lower levels
of support for him in counties with higher
percentages of Catholics, adherents to other
conservative denominations, and those with
dense populations of African Americans.
These ndings are consistent with research
showing the tendency among many Catholics
to switch back and forth between support-
ing the of cial stance of the Catholic Church
and voting for Democratic candidates (Welch
and Leege 1991). More broadly, the ndings
comport with survey data on the topic, par-
ticularly that Protestants were signi cantly
88
Smith, DeSantis, and Kassel
State and Local Government Review
more likely than non-Protestants to support
the bans on gay marriage and that those who
supported the ballot measures were more
likely to vote for Bush (Donovan, Tolbert,
and Smith 2005).
Regarding voter turnout, county-level sup-
port for the measures in the two states does
not appear to be linked to more citizens com-
ing out to vote. The aggregate-level ndings
show that turnout in Ohio was not higher in
counties with greater support for Issue 1, and
in Michigan, turnout was lower as support
for Proposal 2 increased. Counter to some
postelection claims, there is no evidence from
either state that counties with higher propor-
tions of evangelicals, rural voters, or even sup-
porters of the gay marriage bans had higher
turnout levels than other counties. Turnout in
both states, then, increased across the board,
irrespective of support for the anti-gay mar-
riage amendments.
The strategy of using polarizing ballot
measures as partisan cues may rest in part
upon the assumption that potential voters
are willing to politicize what are considered
to be core issues (Smith and Tolbert 2004;
Nicholson 2005). With evangelical leaders
as the driving force behind Issue 1 in Ohio,
this rationale seems to have been the case,
as counties with dense populations of evan-
gelicals supported not only the ballot mea-
sure but also the president. In Michigan,
although counties with heavy populations of
evangelicals and Catholics supported Pro-
posal 2, there appears not to have been an
attendant priming effect for Bush. Perhaps
this outcome was due to the fact that Bush
did not campaign on the issue in Michigan
nearly as much as he did in Ohio. As for the
fact that there is no evidence in either state
that counties with higher levels of support for
same-sex marriage bans and denser evangeli-
cal populations had higher turnout relative
to other counties, scholars of religion and
politics have noted that there are good rea-
sons why socially conservative issues, such as
the anti-gay marriage ballot measures, may
not mobilize all evangelicals equally (Jelen
1994; McCann 1997). While in both states,
same-sex marriage bans on the 2004 ballots
likely primed Bush’s share of the presidential
vote, the electoral impact of the measures was
not universally attributable to the support of
counties with denser evangelical populations,
as the data from Michigan suggest.
Daniel A. Smith is an associate professor of po-
litical science at the University of Florida. He has
published widely on the politics and processes of
direct democracy. He is coauthor (with Caroline J.
Tolbert) of Educated by Initiative: The Effects
of Direct Democracy on Citizens and Po-
litical Organizations in the American States
(University
of Michigan Press 2004).
Matthew K. DeSantis is a Ph.D. student at
the University of Florida. His research interests
include analysis of electoral systems and social
movements. His dissertation focuses on applying
and expanding the theory of electoral capture to
the relationship between the Christian Right and
the Republican Party.
Jason Kassel is a doctoral candidate in political
science at the University of Florida. His disserta-
tion is a historical analysis of the growth of the
U.S. Congress and the American state from 1783
to 1868.
Notes
1. State legislatures in Georgia (76 percent yes), Ken-
tucky (75 percent yes), Mississippi (86 percent yes),
Oklahoma (76 percent yes), and Utah (66 percent
yes) placed referendums on the ballot. Initiatives
were on the ballot in Arkansas (75 percent yes),
Michigan (56 percent yes), Montana (67 percent yes),
North Dakota (73 percent yes), Ohio (62 percent
yes), and Oregon (57 percent yes).
2. There are two different ways to model hypotheses
2 and 3. The rst (results not shown) considers the
dependent variable as a dynamic change value, spe-
ci cally percent change from 2000 to 2004 in turnout
(model 2) and percent change from 2000 to 2004
in vote for Bush (model 3). This type of modeling
leads to the problem of regression toward the mean
because the dependent variable contains information
89
Vol. 38, No. 2, 2006
Same-Sex Marriage Ballot Measures
that should be on the right side of the equation.
Alternatively, both hypotheses can be considered
as panel models, as they include a lagged value of
the dependent variable. The models seek to isolate
the effect on a variable measured at T2 (e.g., same-
sex marriage measures on the ballot), holding ini-
tial conditions (T1) constant. This second approach,
used here, offers a much more conservative test of
the impact of the anti-gay measures, as the lagged
T1values (2000 turnout in model 2; 2000 Bush vote
in model 3) are huge forces that drive T2 in both
models.
3.
The unit of analysis for these models is state coun-
ties. Due to heteroskedasticity concerns, weighted
least-squares regression is used to account for the
wide variation in county size and relative in uence
on statewide electoral results. The weighted variable
is the 2000 population of a county.
4. Measurement of evangelical Protestants is a subject
ripe for academic debate. Following Steensland et al.
(2000), a collapsed denominational typology measur-
ing evangelical Protestants, Catholics, other conser-
vative denominations, and liberal denominations/no
religious adherence is used because it is compatible
with aggregate-level data. For further debate sur-
rounding religious measurement, see Wilcox (1992),
Layman (2001), Jelen (1989), and Kellstedt et al.
(1996). The proportions of evangelical Protestants,
Catholics, other conservative denominations, and
liberal denominations/no religious adherence per
county in Ohio and Michigan were calculated using
census data from the Glenmary Research Center’s
Religious Congregations and Membership in the United
States: 2000 (Jones et al. 2002). The methodology
and actual codes for the religious denominations are
available from the authors.
5. Alternative speci cations for county percent vote
for the Republican gubernatorial candidate or the
Republican Attorney General candidate in the 2002
contests in both states had no statistical or substan-
tive differences.
6. Unemployment statistics for 2004 are from the Mich-
igan Of ce of Labor Market Information, Bureau of
Labor and Economic Growth & Strategic Initia-
tives, and the Ohio Department of Job and Family
Services, Bureau of Labor Market Information. The
models were run using both a static unemployment
variable, which measured the rate of unemployment
in a county in 2003, and the dynamic variable mea-
suring the rate of change in unemployment from
2000 to 2003. There is no predictive advantage in
using one measure over the other, but there is greater
theoretical value in utilizing the dynamic unemploy-
ment variable, as it helps differentiate between coun-
ties that suffered greater increases in unemployment
during the rst term of the Bush administration and
those with chronically high unemployment rates.
7. In two counties (Coshocton and Van Wert) in Ohio,
Issue 1 actually had more total votes cast (for and
against) than the total votes cast for all the presi-
dential candidates on the ballot.
References
Abramowitz, Alan. 2004. Terrorism, gay marriage, and
incumbency: Explaining the Republican victory in
the 2004 presidential election. The Forum 2, no. 4.
www.bepress.com/forum/vol2/iss4/art3.
Belli, Robert, Michael Traugott, Margaret Young, and
Katherine McGonagle. 1999. Reducing vote over-
reporting in surveys: Social desirability, memory fail-
ure, and source monitoring. Public Opinion Quarterly
63:90–108.
Bowler, Shaun, and Todd Donovan. 1998. Demanding
choices: Opinion, voting, and direct democracy. Ann Arbor:
University of Michigan Press.
Branton, Regina. 2003. Examining individual-level vot-
ing behavior on state ballot propositions. Political
Research Quarterly 56:367–77.
—————. 2004. Voting in initiative elections: Does
the context of racial and ethnic diversity matter?
State Politics and Policy Quarterly 4:294–317.
Burden, Barry. 2004. An alternative account of the 2004
presidential election. The Forum 2, no. 4. www.bepress
.com/forum/vol2/iss4/art2.
Button, James W., Barbara A. Rienzo, and Kenneth D.
Wald. 1997. Private lives, public con icts: Battles over
gay rights in American communities. Congressional
Quarterly Press: Washington, DC.
Campbell, David E., and J. Quinn Monson. 2005. The
case of Bush’s re-election: Did gay marriage do it?
Paper presented at the annual meeting of the Mid-
west Political Science Association, Chicago, IL.
Carmines, Edward G., and Geoffrey C. Layman. 1997.
Value priorities, partisanship, and electoral choice:
The neglected case of the United States. Political
Behavior 19:283–316.
Cooperman, Alan, and David S. Broder. 2004. Churches
key to Michigan ght on gay marriage. Washington
Post, October 28, A-6.
Cooperman, Alan, and Thomas Edsall. 2004. Evangeli-
cals say they led charge for the GOP. Washington Post,
November 8, A-1.
Cummings, Jeanne. 2004. How Bush camp won Ohio.
Wall Street Journal, November 19, A-4.
Dao, James. 2004. Flush with victory, grass-roots cru-
sader against same-sex marriage thinks big. New York
Times, November 26, A-28.
Davies, Frank. 2004. Bush’s clear vision aided reelection
bid, Rove says. Miami Herald, November 10, A-1.
Donovan, Todd, and Shaun Bowler. 1998. An overview
of direct democracy in the American states. In Citi-
zens as legislators, ed. Shaun Bowler, Todd Donovan,
and Caroline Tolbert. Columbus: Ohio State Uni-
versity Press.
Donovan, Todd, Caroline Tolbert, and Daniel Smith.
2005. Do state-level ballot measures affect presiden-
tial elections? Paper presented at the annual meet-
ing of the American Political Science Association,
Washington, DC.
Donovan, Todd, James Wenzel, and Shaun Bowler.
2000. Direct democracy initiatives after Romer. In
The politics of gay marriage, ed. Craig Rimmerman,
90
Smith, DeSantis, and Kassel
State and Local Government Review
Ken Wald, and Clyde Wilcox. Chicago: University
of Chicago Press.
El Nasser, Haya. 2004. For political trends, think mic-
ropolitan. USA Today, November 23, A-1.
Foladare, Irving. 1968. The effect of neighborhood on
voting behavior. Political Science Quarterly 83:516–
29.
Freedman, Paul. 2004. The gay marriage myth. Slate,
November 5. www.slate.msn.com/id/2109275/.
Giles, Michael. 1977. Percent black and racial hostility:
An old assumption reexamined. Social Science Quar-
terly 58:848–65.
Giles, Michael, and Arthur Evans. 1986. The power
approach to intergroup hostility. Journal of Con ict
Resolution 30:469–86.
Green, John C. 1994. The Christian right and the 1994
elections: A view from the states. PS: Political Science
and Politics 28:5–8.
Haider-Markel, Donald, and Kenneth Meier. 1996. The
politics of gay and lesbian rights: Expanding the
scope of the con ict. Journal of Politics 58:332–49.
Hero, Rodney. 1998. Faces of inequality: Social diversity
in American politics. New York: Oxford University
Press.
Hero, Rodney, and Caroline Tolbert. 1996. A racial/
ethnic diversity interpretation of politics and policy
in the states of the U.S. American Journal of Political
Science 40:851–71.
Hillygus, Sunshine, and Todd Shields. 2005. Moral issues
and voter decision making in the 2004 presidential
election. PS: Political Science and Politics 38:201–10.
Huckfeldt, Robert, Eric Plutzer, and John Sprague. 1993.
Alternative contexts of political behavior. Journal of
Politics 55:365–81.
Jackman, Simon. 2004. Same-sex marriage ballot initia-
tives and conservative mobilization in the 2004 elec-
tion. Paper presented at the Institute for Research
in the Social Sciences, Palo Alto, CA.
Jelen, Ted G., ed.. 1989. Religion and political behavior in
the United States. New York: Praeger.
—————. 1991. Politicized group identi cation: The
case of fundamentalists. Western Political Quarterly
44:209–19.
—————. 1994. Protestant clergy as political leaders:
Theological limitations. Review of Religious Research
36:23–44.
Jelen, Ted G., and Clyde Wilcox. 1992. The effects of
religious self-identi cation on support for the new
Christian right: An analysis of political activists.
Social Science Journal 29:199–210.
Jones, Dale, Sherri Doty, Clifford Grammich, James
Horsch, Richard Houseal, Mac Lynn, John Mar-
cum, Kenneth Sanchagrin, and Richard Taylor Jones.
2002. Religious congregations and membership in the
United States 2000. Nashville: Glenmary Research
Center. CD-ROM.
Kellstedt, Lyman, John Green, James Guth, and Cor-
win Smidt. 1996. Grasping the essentials: The social
embodiment of religion and political behavior. In
Religion and the culture wars, ed. John Green, James
Guth, Corwin Smidt, and Lyman Kellstedt. Landam,
MD: Rowman and Little eld.
Key, V. O. 1949. Southern politics in state and nation. Knox-
ville: University of Tennessee Press.
King, Gary. 1997. A solution to the ecological inference prob-
lem: Reconstructing individual behavior from aggregate
data. Princeton, NJ: Princeton University Press.
Korte, Gregory. 2004. Gay issue may energize GOP.
Cincinnati Enquirer, September 14, A-1.
Layman, Geoffrey. 2001. The great divide: Religious and
cultural con ict in American party politics. New York:
Columbia University Press.
Lewis, Gregory B. 2005. Same-sex marriage and the
2004 presidential election. PS: Political Science and
Politics 38:195–200.
Lewkowicz, Michael. 2006. The effectiveness of elite
cues as heuristics in proposition elections. American
Politics Research 34:51–68.
Lochhead, Carolyn. 2004. Gay marriage: Did issue help
re-elect Bush? San Francisco Chronicle, November
4, A-1.
Lupia, Arthur. 1994. Shortcuts versus encyclopedias:
Information and voting behavior in California in-
surance reform elections. American Political Science
Review 88:63–76.
McCann, James. 1997. Electoral choices and core value
change: The 1992 presidential campaign. American
Journal of Political Science 41:564–83.
Nicholson, Steven P. 2005. Voting the agenda: Candidates,
elections, and ballot propositions. Princeton, NJ: Princ-
eton University Press.
O’Connell, Sue. 2006 The money behind the marriage
amendments. The Institute on Money in State Poli-
tics. www.followthemoney.org/press/reports/.
Oliver, Eric, and Tali Mendelberg. 2000. Reconsider-
ing the environmental determinants of white ra-
cial attitudes. American Journal of Political Science
44:574–89.
Putnam, Robert D. 1966. Political attitudes and the
local community. American Political Science Review
60:640–54.
Reichley, A. James. 1986. Religion and the future of Amer-
ican politics. Political Science Quarterly 101:23–47.
Rozell, Mark J., and Clyde Wilcox. 1996. Second com-
ing: The strategies of the new Christian right. Politi-
cal Science Quarterly 111:271–94.
Shapiro, Walter. 2004. Presidential election may have
hinged on one issue. USA Today, November 4, A-9.
Shibley, Mark A. 1998. Contemporary evangelicals:
Born-again and world af rming. Annals of American
Academy of Political and Social Science (July): 67–87.
Smidt, Corwin. 1988. Evangelicals within contempo-
rary American politics: Differentiating between fun-
damentalist and non-fundamentalist evangelicals.
Western Political Quarterly 41:601–20.
Smith, Mark. 2001. The contingent effects of ballot
initiatives and candidate races on turnout. American
Journal of Political Science 45:700–706.
91
Vol. 38, No. 2, 2006
Same-Sex Marriage Ballot Measures
Smith, Daniel A., and Caroline J. Tolbert. 2001. The
initiative to party: Partisanship and ballot initiatives
in California. Party Politics 7:781–99.
—————. 2004. Educated by initiative: The effects of
direct democracy on citizens and political organizations in
the American states. Ann Arbor: University of Michi-
gan Press.
Steensland, Brian, Jerry Park, Mark Regnerus, Lynn
Robinson, W. Bradford Wilcox, and Robert Wood-
berry. 2000. The measure of American religion:
Toward improving the state of the art. Social Forces
79:291–318.
Teixeira, Ruy. 2004. Public opinion watch—November
10, 2004. Center for American Progress. www.
americanprogress.org/site/pp.asp?c=biJRJ8OVF&b
=246518.
Tolbert Caroline J., and Rodney Hero. 2001. Facing
diversity: Racial/ethnic context and social policy
change. Political Research Quarterly 54:571–604.
Tolbert, Caroline J., and Daniel A. Smith. 2005. The
educative effects of ballot initiatives on voter turnout.
American Politics Research 32:1–27.
Tolbert, Caroline J., John Grummel, and Daniel A.
Smith. 2001. The effect of ballot initiatives on voter
turnout in the American states. American Politics Re-
search 29:625–48.
Wald, Kenneth D., Dennis E. Owen, and Samuel S. Hill
Jr. 1990. Political cohesion in churches. Journal of
Politics 52:197–215.
Welch Michael R., and David C. Leege. 1991. Dual ref-
erence groups and political orientations: An exami-
nation of evangelically oriented catholics. American
Journal of Political Science 35:28–56.
Wilcox, Clyde. 1989. The new Christian right and the
mobilization of the evangelicals. In Religion and po-
litical behavior in the United States, ed. Ted G. Jelen.
New York: Praeger.
—————. 1992. God’s warriors. Baltimore: Johns Hop-
kins University Press.
—————. 2000. Onward Christian soldiers: The religious
right in American politics. Westview Press: Washing-
ton, DC.
Williams, Vanessa. 2004. Ohio gay marriage initiative
roils skeptics: Some see conservatives pushing bans
in close states to spur turnout for Bush. Washington
Post, May 10, A-5.
Witt, Stephanie L., and Suzanne McCorkel. 1997. Anti-
gay rights: Assessing voter initiatives. Westport, CT:
Praeger.
